Molybdenum price predictions for 1, 5 and 10 years out
August 19, 2012 2 Comments
Writing a blog about molybdenum might arguably be seen as a rather obscure use of one’s time, so why do it?
Molybdenum, in the broader sense, constitutes a topic that is large enough to be interesting, while being narrow enough that one has a shot at getting a genuine understanding — a chance to develop expertise. And molybdenum isn’t trendy. I parachuted into the field because I liked the sound of the word. So there’s a chance to come in ahead of the curve.
Molybdenum, furthermore, is an eight, uhh make that a five, billion dollar industry, yet it’s not out of the question that its price can be predicted with a non-negligible degree of confidence. Molybdenum has fewer moving parts than corn, gold or oil. If the underlying variance stemming from forecasting the general direction of the economy can be hedged away, then molybdenum’s performance is even more amenable to prediction.
My current predictive model is pretty simple. I take an 18-year time series of molybdenum spot prices, and sample at 2-month intervals. Convert these samples to percentage moves. Using bootstrap resampling, draw (with replacement) from the aggregate of percentage moves to generate a forward time series. Repeat 10,000 times.
The results of this model are shown in the plot below. The distribution of outcomes sampled a year from now is shown in red, the distribution five years out is shown in blue, and the distribution a decade from now is shown in black. Both the blue and the black curves have substantial tails at high price, extending well off the 50 USD/lb x-axis limit. A year from now, the 10,000 trials average to 12.45 USD/lb, with a standard deviation of 4.75 USD/lb. Roughly speaking, the model suggests that there’s a 15% chance that a year from now, the price of mol will be above 17 dollars per pound, and a 15% chance that it’ll be under 8 USD/lb, an outcome that wouldn’t be too rosy for shares of GMO.
Remember, though, it’s just a model, and an excruciatingly simple one at that. Next up, some analysis of the model with an eye toward improving it.